CONTENTS
Pages
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Introduction
1
..........................................3
Development of WRIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
......................................
Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
System Design
UsersofWRIS
4
5
6
......................................... 6
....................................
Map Data Compilation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6
Map Data Manipulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7
Capabilities of WRIS
7
................................. 8
Cost Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Digitizing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Labeling and Editing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Data Manipulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
System Maintenance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
FutureOutlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10
Production Information
Literature Cited . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
The Authors
I
are assigned to tile Station's research unit investigating measurement and
analysis techniques for management planning, wit11 headquarters in Berkeley, Calif. ROBERT M. RUSSELL, the unit computer programmer, received a bachelor's degree in mathematics a t the University of Michigan
(1956), and worked as a programmer at the University of California,
Berkeley, from 1958 u n t i 1966, when he joined the Station staff. DAVID
A. SHARPNACK, a research forester, was educated at the University of
Idaho (bachelor's degree in forestry, 1961) and the University of California, Berkeley (master's degree in statistics, 1969), and has been with the
Station since 1962. ELLIOT L. AMIDON is in charge of the measurement
and analysis techniques research unit. He earned a bacl~elor's degree in
forest management at Colorado State University (1954) and a master's
degree in agricultural economics at the University of California, Berkeley
(1961), and was assigned to production economics research at the Station
u n t i he assumed his present position in 1971.
ACKNOWLEDGMENTS
The Wildland Resource Information System (WRIS) was developed
by the Station's research unit investigating measurement and analysis
techniques for management planning. WRIS was developed primarily
for the Branch of Management Plans and Timber Inventories, California
Region, Forest Service, U. S. Department of Agriculture, San Francisco.
We gratefully acknowledge the help of the Branch staff; Klaus H.
Barber, inventory supervisor, Stanislaus National Forest; and E. Joyce
Dye, computer programmer with the Station researcll unit staff.
At various stages of system development, we used data from the
following National Forests in California: Stanislaus, Eldorado, and
Sierra.
SUMMARY
Russell, Robert M., David A. Sharpnack, and Elliot L. Amidon
1975. WRIS: a resource information system for wildland management. USDA Forest Sew. Res. Paper PSW-107, 12 p., illus. Pacific
Southwest Forest and Range Exp. Stn., Berkeley, Calif.
Oxford: 624:U681.3:(084.3)
Retrieval Terms: timber management; wildland management; resource
use planning; computer programs; map compilation; WRIS; management information systems.
The Wildland Resource Information System
(WRIS) is an operational computerized procedure for
acquiring spatial data for management planning.
Though designed primarily for timber management
use, it has application for land-use planning in general. WRIS is a production tool for which detailed
instructions are provided for both manual and computer operations. The system is intended for continuous use by institutions rather than individuals. It
is designed to process a workload of 200 to 400 maps
a year with a staff of two or three people at a central
location, such as a regional office or corporate headquarters. Computation is performed on a mediumsized computer, the UNIVAC 1108, over a high-speed
batch terminal. Other mandatory equipment needs
include access to a scanning device for automatic digitizing, and an incremental line plotter. A hand digitizer is optional, but highly desirable if source maps
vary from simple to highly complex. Software consists of over 20,000 standard FORTRAN statements
compiled by UNIVAC's EXEC 2 Executive System.
The publication Wildland Resource Information
System: User's Guide (USDA Forest Service General
Technical Report PSW-10) details the operating procedures for using WRIS com.puter programs and
digitizing hardware.
WRIS provides a means of collecting, processing,
storing, retrieving, updating, and displaying geographic data, and makes possible the performance of logical
operations on these data. System capabilities include
data reduction from maps or orthophotographs at
varying scales, printed tabular and plotted graphic
display, computer-aided manual editing procedures,
and a broad range of data-manipulating features.
Logical operations include merging and overlaying of
map Wes, selection of polygons with measurement of
their area and perimeter, and extraction of rectangular subsets of data withii a map border.
The basic data collection unit in WRIS is the polygon-a plane figure consisting of vertices connected
by l i e segments. Each polygon has a unique number
for computer processing and a category such as "red
fir, volume category.. .," which has meaning for inventory purposes. This data structure differs from a
more common one in which the squares of a f x e d
grid are labeled. A few other large-scale systems have
a similar approach, which minimizes storage but does
require more complex data-manipulating algorithms.
About 10,000 polygons will cover a National Forest
and span 1000 different categories. We report here
statistics drawn from experience with three California
National Forests because they affect system design
and have implications for management planning.
Modifications and extensions of the system described here are inevitable because of the continual
flow of new teclmology. Software conversion to the
EXEC 8 version of FORTRAN is already underway.
Adaptation to IBM 360-type computers is planned,
and conversion to PL/l is an attractive possibility.
Several editing and updating alternatives are available,
but more will be tested because these are key procedures in the cost of map data reduction.
Finally, we will seek ways to achieve communication between this and other geographic information
systems. We need data format con~patibility, standardized terminology, and exchange of methodology
before wildland resource information can be routinely available to forest managers.
T
he rapidly rising value of wildland goods and
services in recent years has put increasing pressure on wildland managers to make the most of
tile resources under their control. "The Environmental Program for the Future" (U.S. Forest Service
1974) provides some measure of the size and scope of
the land management job facing the U.S. Forest Service. In order to provide in the next 5 years the same
proportion of goods and services it has in the past, it
"must conduct silvicultural examinations and prepare
prescriptions for 19 million acres; carry out inventories on 50 forests covering 31 million commercial
forest acres and prepare timber management plans on
61 forests covering 37 million commercial forest
acres." These are major tasks facing those responsible
for the proper management of the National Forests.
What kinds of information and capabilities are
needed to develop an acceptable management plan?
Barber (1973) suggests six requirements:
1. The location and acreage of all stands on the
Forest.
2. A means of imposing physical and administrative constraints on the forest practices that will be
allowed in individual stands.
3. A description of all stands, covering their
structure, composition, volume, growth, and yield.
4. A means of predicting what would happen to
each stand if it were harvested or, alternatively, if it
were left to grow.
5. A set of techniques for analyzing the Forest's
growing stock and land-use patterrrs to produce a
management plan.
6. A means of updating tile plan to accommodate changes in land base, stand structure, and
management goals.
Of the six requirements, the collection and manipulation of geographic data can be d ~ most
e
costly and
time-consuming steps in preparing a management
plan. The Wildland Resource Information System
(WRIS) was developed to handle this kind of problem.
T l ~ eWRlS user obtains the location and acreage of
a l l stands on the forest by digitizing maps of forest
types. Unlike most earlier map systems, WRIS digitizes the maps by using a scanningmicrodensitometer
and two computer programs: FREQTB converts the
scanner output to a binary map; POLLY extracts
stand boundaries, which we call polygons, from the
binary map, and produces a list of all stand areas and
a magnetic tape containing all the stand boundaries.
Constraints are imposed on managing the forest in
two steps using WRIS. First, maps delineating areas
which impose either pllysical or administrative constraints on management are digitized in the same way
as the stand maps. Second, by using the output for
the stand map and the constraint map, a computer
program (MOSAIC) overlays the two maps to produce polygons which represent a combination of information on forest types and constraints. The acreages of
these polygons are used, along wit11 other data, in the
analysis necessary to develop amanagement plan.
Updating is easily handled by WRIS. Changes in
either the forest-type or the constraints layer can be
made. MOSAIC can be rerun and dle new acreages
entered into the analysis. WRIS can plot tlle original
and the overlaid maps by t l ~ eprogram CHART, or
provide various useful lists by two programs (GOSSIP
and RUMOR). These outputs from WRIS are particularly useful for developing a work plan to carry out
dle new management plan.
An important feature of WRIS is that it is not
limited to the scquence of steps just described for
developing a management plan. The capabilities of
WRIS can be used for any process that begins with
more than one layer of maps and combines these
layers to provide the data for decisionmaking.
This paper describes the development of the Wildland Resource Information System and its capabilities
and characteristics, provides production information
about using the system, and considers the outlook for
future modifications.
The publication Wildlarid Resource I~zfoormatioiotz
System: User's Guide (USDA Forest Service General
Technical Report PSW-10) details the operating procedures for using WRIS computer programs and digitizing hardware. Copies of the report and the computer programs are available on request to: Director,
Pacific Southwest Forest and Range Experiment Sta-
tion, P.O. Box 245, Berkeley, California 94701, Attention: Computer Services Librarian. The programs
will be copied on a magnetic tape, t o be supplied by
the requestor. Before sending the tape, the requestor
should contact the Computer Services Librarian,
giving the tape format desired.
DEVELOPMENT OF WRIS
The potential usefulness of a computerized geographic information system was recognized years ago.
But the development of such a system has been
hampered by three major obstacles: First, data collection was limited to hand digitizing technology;
second, the computational methods available required
data to be cateeorized in a cellular fashion: and third.
it was not clear that a system that solved technical
problems would also be economically feasible.
Once the need for a geographic information system was recognized, the next step was to establish its
scope. This meant considering the degree of centralization, deciding which functions and decisionmaking
levels would be assisted, and establishing a production
rate in balance with the expected annual map requirements. Compromises had to be made, and the result is
considered desirable rather than optimal.
In a highly decentralized organization, such as the
U.S. Forest Service, a configuration usable at the
National Forest headquarters level may be desirable.
Yet the economy of scale yielded by scanner digitizing suggests centralized processing at the regional or
multiregional level. Hence, the role of hand digitizing
has diminished for economic rather than technical
reasons.
WRIS is functionally oriented toward timber in
order to minimize the cost of map data reduction, as
timber data was most easily available. Any other wildland data, such as geologic and soil classifications,
that can be represented on maps, can be handled by
the same procedures as timber data. New procedures
can be improvised to adjust the system to fairly
drastic changes in the input format. Point data such
as elevations, or lineal features like roads, might warrant special treatment. We expect the bulk of the
work to consist of identification of areas with irregularly shaped but connected boundaries, and the scope
of the system reflects this fact.
Within the Forest Service's California Region,
timber management planning alone requires about
300 new maps per year. Half of these maps are con-
-
cerned with biological data-largely timber types.
Other management activities collectively generate at
least another hundred maps. A minimum processing
rate is therefore two sheets per day.
System Design
System design requires the matching of the user's
needs to the available resources. The client should be
clearly identifiable with definable objectives. This
ideal is difficult to attain because there are usually
many potential users. I t can be approached by selecting the appropriate design strategy. A common one is
to conduct surveys of user's information requirements-a method requiring substantial resources
(Boeing Computer Service, Inc. 1972; Raytheon Co.,
Autometric Operations 1973). An alternative, when
the designers are already familiar with the user's problems, is to make simplifying assumptions and then
provide an information system which seems just adequate. Then the system can be used to reveal the
user's decision process. Later, portions of the system
can be upgraded incrementally and performance continually evaluated. We took this approach in the design of WRIS.
A major influence on our design was control over
the map compilation process, following aerial photography. To allow both automatic scanning as well
as hand digitizing, we made slight changes in map
annotations and in line widths in map drawings.
Another influence on design was the resolutionfmeness of detail-required to distinguish lines in
aerial photographs. We could place forest-type lines
effectively within about 50 feet (15 m) of true
ground location. The inaccuracy represented by this
distance is not critical, as it would be in depth soundings on navigational charts, for example. The appropriate, much less optimum, resolution could be
defmed only by substantial analysis.
A further influence on our design of WRIS was
the several sources of complexity found in forest
r
maps. One source is the characteristic of lines.
Straight lines allow data to be compressed because
essentially only changes in direction must be recorded. That fact is undoubtedly helpful in urban geographic information systems. But natural phenomena
show few regularities; the only straight lines we encountered were administrative boundaries. The foresttype maps were quite detailed. Polygons are plane
figures consisting of three or more vertices connected
by line segments. The smallest polygon allowed on a
map was 5 acres-about 10,000 polygons occurred on
a Forest.
To express the attributes of a polygon, we use a
label. For each polygon, its coordinate position is
recorded and a label assigned. A label consists of from
1 to 36 characters. It is not unique, but may occur
many times on a map-once for every polygon with
ille same attributes.
To describe the attributes of the 10,000 polygons
recorded on a National Forest, we needed about 1250
labels to show classifications of forest type. Perhaps
the classifications were too fine: almost any management planning would require about a twenty-fold
reduction in the number of categories. Yet detail
should be in excess of immediate needs. A suggested
strategy is to have low-access storage of the original
detail so the composition of the groups can be
changed. This additional flexibility is worth the additional setup cost.
Clearly, availability of hardware can affect system
design. The ubiquitous line printer, designed for
printing characters, has served in lieu of expensive
graphic output devices since the early 1960's (Thornburn and others 1973). WRIS originated in a multipurpose batch processing environment. The batch
operation differs significantly from other installations
in that the terminal can communicate easily with
varying brands of distant computers. The peripheral
graphic devices, digitizers and plotters are singlepurpose equipment.
Software availability had little effect on system
design. Tile few computer programs described in the
literature were coded mainly by geographers (Tobler
1970). This lack of programs simply reflects the state
of the art. There is still uncertainty and disagreement
as to the correct method for storage and manipulation of graphic data. The "best" ways change rapidly
with new developments in computer technology. We
can expect algorithms as well as devices to be continually redesigned.
All large-scale geographic information systems
require specialized equipment for map digitizing and
graphic display. Often a sophisticated device must be
developed before technical feasibility can be demonstrated. A prime example is the Canadian Geographic
Information System, which uses a drum scanner designed expressly for the Canada Land Inventory
scribed map sheets (Tomliison 1967). Another
Canadian system, on a smaller scale, perfoms cartographic manipulations without access to a shared
computer. The configuration includes a storage display connected to a minicomputer by a specially
designed interface (Graphic System Design and Applications Group 1972).
Our approach was to use commercially available
services and to avoid hardware modification. For
example, the hand digitizer selected was unusual
because it did not have any moving parts connected
to its cursor, and it was obtained with as few optional
"extras" as possible. As a result, this first production
model is still in use. All adaptations to changing
technology are by means of software.
Maintaining flexibility through software tends to
rule out the minicomputer in favor of medium or
large computers shared by many users. AIJ development has taken place on a UNIVAC 1108 computer,
but conversion to other computers is feasible. The
goal of flexibility has also resulted in a modular
program structure, in FORTRAN. Similarly, the use
of assembly languages or FORTRAN language features peculiar to one computer configuration has
been avoided as much as possible. A large production
rate may warrant some software specialization, however, to minimize input-output cost.
Testing
Testing has involved the repeated application of
two steps. First, the requisite computer programs
were debugged separately and as a group by using
sample data. Then a series of about 50 maps or one
forest overlay was processed. Each repetition suggested changes for the next. The first group of maps
caused the most substantial change, from a process
centered on hand-digitizing to one relying on an automatic scanner. About 236 township maps, covering
two National Forests, have been completely processed.
Because technology is always changing, testing for
technical feasibility will follow each change. But the
final test, that of economic feasibility, cannot be made
in a researcll and development environment. Only users
can provide the stable working conditions needed to
rationalize work flow. Cost data should be based on a
system operated routinely by experienced personnel.
These data should be collected during a year of routine
work after the system llas been implemented.
Implementation
The major consideration before system installation
is the expected volume of work. The computer programs can be compiled and stored for such a small
cost that even infrequent use would justify them.
But, even for such use, at least one person would be
required t o operate specialized equipment and assist
with occasional complex editing problems. A clearly
appropriate workload would be about 200 maps per
year processed by a two-person staff.
The source maps are generally prepared under
contract, as most of the data processing could be.
Within the Forest Service, we recommend that polygon identities be encoded at Forest headquarters,
where the resource photograpl~yis available to correct erroneous map symbols. All subsequent work can
be performed at the Regional level or commercially.
USERS OF WRIS
Most users of this information system are the
management planners in a decentralized organization.
This practice implies a two-way flow of data between
at least two management levels, such as Forest and
Region headquarters. Current users include timber
management
planning- staff members, who rewire
facts about Ole resource, such as the acreage distribution of forest types, the timber volumes and the
growth rates associated with them. In turn, these data
are interpreted, and in accordance with policy guidelines, are passed forward t o successive decisionmakmg
levels. Often information is prepared as computer
input for subsequent analysis by linear or dynamic
programming. The formal linking of computerized
spatial information and mathematical models has already begun. For example, the combining of "TRI"
and "RAM" for compartment management has been
described in the report on The Siuslaw Mode1,U.S.
Forest Service, Pacific Northwest Region 1973).
The kinds of map information used vary by location in the United States. On the Stanislaus and
Eldorado National Forests in California, for example,
two types of mapped data are essential to estimating
timber inventory. Timber-type maps are prepared
from large-scale (1: 15,840) resource photography in a
General Land Survey, township format. Together
with delineations of administrative boundaries, these
maps provide spatial information on management
possibilities. Next, a management component map is
prepared, showing land classifications that will affect
harvesting or cultural activities. The combination of
tbese two maps, with corresponding volume and
growth data, provides the physical information for
computing.timber management
alternatives.
.
It is easy to visualize how information from other
functions or disciplines could be applied. Soil information could be incorporated as an additional stratum affecting cultural practices. At least three layers
of information would be needed for a single function
such as timber management; multidisciplinary problems might require several times that amount.
Some shifts can be anticipated in the needs of the
users as well as in graphic technology. As demands on
the forest resource increase, so will the need for more
detailed "in-place" information. Conventional resource pllotography and mapping may be complemented or replaced by orthophotography. Source
negatives may be taken by high-altitude aircraft or
satellite. Presumably, classifications for various purposes will still be delineated from this source material, and for that reason few modifications to the
existing system will be needed. Large changes will
arise when the user expects analytic additions to the
system and not just acreage calculations and summations.
CAPABILITIES OF WRIS
WRIS provides the means of collecting, processing, storing, retrieving, updating, and displaying "inplace" resource information, and makes possible the
performance of logical operations on it. Two characteristics of these capabilities are notable: First, they
cannot be regarded as an optimal mix because the
multiple objectives of the. land manager, and t l ~ e
many constraints on ids decisionmaking, are too com-
plex to allow such judgment. Second, these capabilities exist in other geographic information systems of
comparable size (Tomlinson 1972); the major difference is the ease with which they can be exploited.
0 Data from maps can be recorded at various
scales-provided a few geograpl~icreference marks are
provided for alinement. The two scales used so far in
processing National Forest data are 4 inches and 2
I
I
inches per mile for the same townships. Two coordinate systems are adequate for California: State Plane
(Lambert Conformal) and geographic (latitude/
longitude) coordinates. Transverse Mercator is included in the system for application in other States,
and Universal Transverse Mercator (U.T.M.) is also
available.
Images of a map can be displayed on a line
printer, but this is intended only for intermediate or
editing purposes. Plots, either for editing purposes or
for final use, are drawn by either flatbed or drum
ink-line plotters of moderate accuracy.
0 Computation is performed on a UNIVAC 1108
Computer in FORTRAN over a high-speed batch
terminal.
0 Data can be manipulated by editing and by
performing logical operations. Editing, the most expensive task, is the hardest to describe because of the
diversity of errors possible. A major problem in editing is controlling the quality of the original map data.
0 Spatial analytic techniques and logical operations include extraction of data on the basis of political or administrative boundaries; statistical summaries
of extracted data, including area and perimeter measurements; merging and overlaying map fdes; and removing a rectangular subset of data within a map
border (windowing).
0 Finally, the system can perform routine bookkeeping tasks. For example, maps and magnetic tapes
must be logged in and out, digitized data packed for
low-access storage, and backup fdes maintained in
case of loss. The procedures allowed for would be
adequate for an expected production rate of 200 or
300 maps a year.
(attributes) appear on the map in blue, which is not
visible in a black-and-white copy negative. The procedure just described is used for the two other types
of maps digitized, except that scales differ and paper
is used rather than plastic for the base.
Our experience with California Foresfs should
help judge t l ~ efeasibility of handling other jobs with
WRIS. Our forest-type maps were quite detailed, with
200 to 300 polygons per map sheet. Maps corresponding to the same areas, showing management
components, had 50 t o 100 polygons per sheet. (The
components are areas suitable for various types of
logging, ownership, use as recreation sites, and similar
purposes.) Finally, a simple map was drawn showing
only administrative boundaries. More information,
particularly soils data, would be needed for timber
management planning, but forest-wide "layers" are
riot now generally available.
Data processing with WRIS can be accomplished
using any map format or scale. A township (36 square
mile) outline at 2 or 4 inches per mile is used in the
California Region. The familiar U.S. Geological Survey quadrangle format would be preferable, however.
The public land sulvey grid is unwieldy because it is
an irregular format. Virtually any mathematically
defined coordinate system would be more convenient; the township outlines could be included for use,
just as any other layer of information could be.
Map-scale variations are handled easily. The scale of
input can be quite large because maps can be reduced
photographically. The only size requirement for the
output is that it be convenient for manual editing.
This size rarely exceeds that of the original map.
Map Data Compilation
Map Data Manipulation
Preparation of forest-type maps often begins with
aerial photography. In the California Region, resource
photography is generally taken in color at a scale of 4
inches per mile (1: 15,840). From stereographic coverage of an entire Forest, type maps are drawn by a
contractor, who uses standard mapping procedures.
Type boundaries down to a minimum of 5 acres are
delineated on the photographs in accordance with
regional defmitions. The lines are rectified and transferred to a plastic map sheet. Width of the black ink
lines is determined by scanning considerations: 0.025
inch is typical for a 2- by 2- foot township map.
Map compilation results in a closed network of
lines, with each polygon identified by a label indicating forest type. (A closed network means that only
polygons are on the map sheet. All lines join other
lines, including the map (township) border). Labels
After the maps have been drawn, the next manual
step is to assign each map polygon a label. Because
errors, such as missing or conflicting identities, must
be corrected, this work is done at the Forest headquarters, wilere the photography is available for reference. The only problem is that maps often include
special characters unknown to the computer. Consequently, a translation scheme must be devised before
labeling is done. The labeling itself is a simple process.
A polygon is selected arbitrarily. I t is identified by a
label, and the information is keypunched. A single x,
y location falling within the polygon is also recorded
and is thereafter associated with the label. Successive
polygons are labeled until all have been recorded.
Some map errors are corrected, and a few labeling
errors are generated-only t o be detected during subsequent steps.
S t e r e o Photo
Interpretation
I
Labels
Digitizer
The labeled and logically correct map is then
ready for digitizing (fig.1). Like labeling, digitizing is
a fairly straightforward process because two digitizers
are available: manual or automatic. Simple maps,
such as administrative boundary sheets, can be digitized manually by hand-held digitizers; more complex
maps are digitized automatically by microdensitometers (fig. 2). For automatic digitizing the map is
first photographically reduced to a black-and-white
negative. Then it is scanned line by line (taster scan),
and the densities are recorded on magnetic tape. The
tape is stored for later processing.
Assembling the raster-scan data into valid lines
and polygons requires editing, a complex process.
Essentially it progresses in steps, with intermediate
tabular and graphic output to display errors. The
number of steps varies because some errors may require recycling back one or more steps. Editing ceases
when the total area is correctly recorded within a
specified tolerance. Editing completes the data reduction procedure except for data packing and copying
for long-term storage.
Once correct, the map data are available for retrieval and manipulation. In forest inventory work,
the usual next step is to lay one map sheet over
another to obtain logical combinations. The output
from each pair of township maps is acreages. These
acreages can be accumulated for an entire forest. This
application alone has many variations, depending on
the immediate problem.
I
0
Scanner
Editing
Plots
0
Manipulation
Figure 1-WRIS produces information from forest-type maps in a
multiple-step process. A map derived from resource photography is
divided into polygons. Each polygon is labeled by recording its coordinate position. A map is either hand-digitized or is automatically
scanned by a microdenritomer. I t is than edited to repair data liner
and to account far all polygons. The data produced can be manipulated and displayed in various graphic forms.
PRODUCTION INFORMATION
Cost Estimates
The average township map can be processed for a
direct cost of about $170. This estimate reflects a
number of underlying assumptions and conditions.
On the basis of past experience in acquiring data for
two National Forests, it assumes two or three trained
people operating at a rate of 200 or 300 map sheets a
year. The major means of digitizing is by the rasterscanning method. Given these various conditions, estimated costs, by functions, are:
Cost per township (dollars)
Process:
Scanning
Labeling
20
20
Editing:
Labor
Computer time
Plots (1%)
Total
15
80
30
165
The manl~ourestimates given are net, in that relaxation periods or miscellaneous interruptions are
I
Figure 2-Digitizing is a key step in preparing maps in
WRIS. Far simple maps, i t can bedone manually with
a hand-held digitizer. left; for more complex jobs,
optical densities are measured by a scanning micro.
densitamer, right.
excluded. The cost of computer time in 1974 was
about $325 per hour. A labor cost of $5.00 per hour
is assumed. All hourly data are based on the performance of six well-trained students over a period of 2
years. As experience has shown that individuals vary
in performance, only averages and "rules-of-tlmmb"
are warranted.
Cost estimation is greatly simplified because we
can ignore fiied or setup costs. All equipment senices, with the possible exception of ascanner, can be
rented. Consequently, the vexing problems of depreciation, salvage value, and the allocation of joint
costs, are not considered here.
Digitizing
-
Automatic recording of map negative densities in
a raster format requires a specialized machine. The
PDS-1000 scanning microdensitometer we use was
originally purchased for digitizing aerial photographs.
Because it is highly accurate, it is unnecessarily slow
for our map work. It would cause a bottleneck if
production were doubled: the typical million-point
scan takes 10 hours. Many other digitizers are now
available and, while these have not yet been tl~oroughly investigated, it is evident that some of the
cruder devices can scan in minutes, or even seconds,
jobs which now require hours.
Labeling and Editing
or two, within three or four passes, for an average of
3 hours of work.
Before labeling can begin an encoding scheme is
developed. The original map designations usually have
characters wl~ichcannot be keypuncl~edor replicated
by a print chain. Once computer-readable equivalents
are memorized, labeling time simply depends on tl1e
number of polygons per map. The time requirement
has varied between % and 6 hours; 4 hours are
needed for a typical map.
Editing requires an ability to visualize spatial relationships, such as is needed to solve jigsaw puzzles.
Fortunately, most development effort has been applied to t l ~ eediting problem, and numerous aids have
been documented.
Editing is divided into two parts: repairing the line
data, and accounting for all polygons. Line repair
begins with printing the scanner output tape, that is,
the 0,l matlix, in strips. This takes 1% to 3%
minutes of computer time, depending on the amount
of "blank" space. A visual scan of the strips detects
gaps in lines, requiring the addition of 1's to fdl tbem
in. Deletions are needed occasionally, as when the
space between closely adjoining lines has been
bridged by tile scanner. Recording corrections takes
about 2 hours and is clearly affected by the quality
and extent of the original ink lines.
During the second phase, the polygon labels prepared earlier are input along with corrections to a
polygon-extracting program. The program checks
incoming labels, computes areas, and reports various
error conditions for editing. The first time through,
the more obvious errors are corrected and the job is
then resubmitted. Later, as errors become more
subtle, a plot may be necessary. This iterative process
continues until the total area is within a prescribed
tolerance of the area enclosed by tlle map border.
Generally the maps are correct, with the aid of a plot
Data Manipulation
Once t l ~ emap data are correct all subsequent
expense depends on the extent t o which the data are
used. The timber inventory work performed is area
oriented. The only operation immediately required in
our experience was the intersection of all polygons on
two or more map layers and the calculation of t l ~ e
polygon areas created. This expense is proportional to
the number of input polygons (fig.3). The demand
for other operations, such as selected overlays and
plots, is expected to increase soon. We can expect the
cost roles of acquiring and manipulating map data to
reverse tl~emselvesas use increases.
System Maintenance
AU cost estimates were based on the assumption
that services were rented. Purchasing equipment, particularly digitizers and plotters, may be a better alternative at a production rate of a few hundred maps per
year. In tllat case, the manufacturer's allowance for
maintenance (often about 10 percent of purchase
price) would be prorated over the expected map
output. Buyers have a wide range of choices. A digitizer-plotter combination, for example, could range in
price from $30,000 to $80,000.
Software will be maintained indefinitely for use
by the National Forests. But available resources may
not permit a continual, general distribution of program modifications. All computer programs have
documentation in the form of comment cards to
minimize maintenance cost. Most users will have to
write software to read digitizer output tapes for the
various brands available.
FUTURE OUTLOOK
Computerized systems are particularly subject to
technological change. Geographic information systems are vulnerable from three directions: (a) graphic
hardware alternatives are changing in line wit11 satellite teclmology; (b) methodology is scarce because
little docu~nentationis available; and (c) applications
for existing systems are quite rudimentary. Evidently
illere will be a lag before users drop their mental
constraints on possible applications. This bar may be
due to past association w ~ t hthe tedious task of extracting graphic information by hand. For example,
manual methods for determining line-of-sight or terrain slope from elevation data have been available for
decades. Yet, only recently have these applications
been computerized (Amidon and Elsner 1968, Sharpnack and Akin 1969).
Software extensions of present capabilities will
take priority over Rardware modifications. The
UNIVAC 1100 Series Executive System (EXEC 8)
version of FORTRAN will replace the present EXEC
2. Conversion to IBM 360 computer system is
planned and PL/I language conversion is a possibility.
10
Figure 3-The cost of overlaying maps in
WRlS is directly related to the amount
I
I
I
I
I
100
200
300
400
500
I
600
Number of polygons
of map detail.
Several National Forest data bases are now undergoing changes. It is inevitable, therefore, that more
effort will be spent on updating. Several alternatives
already exist, such as the overlay procedure, but
others will need development as the frequency of
updates grows. Hand digitizing methods, less attractive than scanning for capturing initial data, should be
useful for modifying local data. Hand digitizing may
also prove attractive for adding point and line data.
Once the National Forest data bases are reliable
and readily available, requests for storage and retrieval will increase. Then the currently moderate
capability of accessing acquired information will
require extension. Perhaps an existing general purpose
system, such as GIM (General Information Management), can be used, thereby minimizing development
costs.
Finally, some means will be found to achieve
communication between geographic information
systems. This tie-in could begin with tape format
compatibility. And eventually, it could evolve into
standardization of terminology and software exchange, as well as other areas signifying maturity in
this expanding field.
LITERATURE CITED
Amidon, Elliot L., and Gary H. Elsncr,
1968. Delineating landscape view areas: a computer approacll. USDA Forest Serv. Res. Note PSW-180, 5 p.,
illus. Pacific Southwest Forest and Range Exp. Stn.,
Berkeley, Calif.
Barber, Klaus.
1973. Inventory and allowable cut calculations. In Perm.
Assoc. Comm. Proc., 1973. West. Far. and Conscrv.
Assoc. Portland, Oreg., p. 170-173.
Bocing Computer Services, Inc,
1972. Natural resource information system.Vols. I to IV.
Scattlc, Wash.
Graphic System Design and Applications Group.
1972. Preliminwy user's notes on interactive display
system for manipulation of cartographic data, 43 p.
Univ. of Saskatchewan Electr. Eng. Dep., Saskatoon,
Canada.
Rayilleon Co., Autometric Operations.
1973. The development of a natural resource information
system. Vols. I to V. U.S. Dcp.. Int. Contract, K5 1C
14200693, Wayland, Mass.
Sharpnack, David A,, and Garth Akin.
1969. An algorithm for computing slope and aspect from
elevations. Photogram. Eng. 36(3):247-248.
'llornburn, G., K. M. Magar, and G. S. Naglc.
1973. An information system for rural land-use plannmng,
BC-X-75.91 p. Can. Dep. of t l ~ cEnv~ron.,V ~ c t o n s B.C.
,
Tobler, W. R.
1970. Selected computer programs, 162 p. Unn. M~chlgan, Dep. Geogr., Ann Arbor, Mlch.
Tomlmson, R. F.
1967. An introduction to Ute geograph~cinfomation
system of t l ~ eCanada land inventory, 23 p. Can. Dep.
Far. and Rural Dev., Ottawa. Canada
Tomlinson, R. F., editor.
data Itandling.
1972. Geomaphical
..
- PTOC.,UNESCOIIGU
Second Symp. on Geogr. I n t Syst., Otta~vs,Canada,
1351 p. IGU Comm. on Gcogr. Data Sensing and Process., Ottawa, Canada.
U.S. Forest Service, Pacific Northwest Region.
1973. The Siuslaw model: a design for compartment
management, 25 p., illus. Portland, Oreg.
U.S. Forest Service.
1974. Environmental program for tllo future. A long
term forestry plan. Washington, D.C.(unpaginatcd).
The Forest Service of the U.S. Department of Agriculture
. . . Conducts forest and range research at more than 75 locations from Puerto Rico to
Alaska and Hawaii.
. . . Participates with all State forestry agencies in cooperative programs to protect and improve the Nation's 395 million acres of State, local, and private forest lands.
. . . Manages and protects the 187-million-acre National Forest System for sustained yield
of its many products and services.
The Pacific Southwest Forest and Range Experiment Station
represeqts the research branch of the Forest Service in California and Hawaii.